1

Combined Reporting Methodology

Shane SandersRevenue Forecasting Analyst

October 2005

Business Tax Reform Commission

• Formed March 4, 2004• Purpose: “The Commission is established to

evaluate the current business tax structure in thisCommonwealth and recommend changes in theCommonwealth's business tax structure that willbroaden those tax bases, thereby allowing ratesto be reduced, leveling the playing field, andcreating a fairer business tax climate.” (ExecutiveOrder 2004-3)

• Primary Goal: Reduce CNIT rate (currently 9.99%)• Proposals must be revenue-neutral and only

affect taxes paid by businesses

2

BTRC Recommendation

• Mandatory unitary combined reporting• CNIT rate reduction from 9.99% to 6.99%• CNIT apportionment – increase sales factor weight

from 60% to 100%• Uncap net operating losses (NOLs) generated by the

combined group• Adoption of market-based sourcing of sales for

apportionment purposes• Adoption of a 1% net income tax on pass-through

entities

BTRC Recommendation

• Expected CNIT in 2005 - $1,700 million• Cost of rate cut (9.99% to 6.99%) - $510 million• In order to drop the CNIT rate but still remain revenue-

neutral, other options put in place– Mandatory unitary combined reporting– 1% entity-level tax on pass-throughs

3

Governor’s Budget Proposal

• Mandatory unitary combined reporting• CNIT rate reduction from 9.99% to 7.99%• CNIT apportionment – increase sales factor weight

from 60% to 100%• Uncap net operating losses (NOLs) generated by the

combined group• Adoption of market-based sourcing of sales for

apportionment purposes

Methodology – Data Sources

• Pennsylvania Data– 138,000 C Corporations in 2000– Data used for:

• Separate company liabilities• Apportionment

• IRS (BMF) Data– Used to compare federal income reported to Pennsylvania vs.

income reported to the IRS• Of the 138,000 C Corps, 63,500 matched the BMF data by both EIN

and income – No CR impact• The remaining 74,500 did not match – Likely CR Impact

4

Methodology – Data Sources

• Minnesota Data– Provided combined income and apportionment data for all

companies filing in Minnesota that apportion– Tax years 1999,2000,2001– Matching the 74,500 affiliated C corporations resulted in the

identification of 6,472 combined unitary groups filing inMinnesota

Methodology – Sample Selection

• 2000 Data– 25,017 corporations in Minnesota that become– 16,788 combined groups of which– 6,472 have parent EINs that match PA data and therefore have a

known separate company federal income value. All other groupsare assumed to have a PA federal income amount of $0.

• Data used to create sample has the followinginformation:– Parent EIN– Parent SIC (from PA data, when available)– Income amounts (separate company and combined group)– Income difference (absolute value)

5

Methodology – Sample Selection

• Sample selection done by looking at the absolute valueof the difference between MN and PA income

• Greater than $1 billion difference– 107 groups have an income difference > $1 B– Due to the large income discrepancies, this group is sampled 100%

• Less than $1 billion difference– 4,751 groups have an income difference < $1 B and > $1 M– 118 of these groups are sampled. Mix is based on SIC and

income difference.• Less than $1 million difference

– 7,371 groups have an income difference < $1 M and > $0– 5 of these groups are sampled.

Methodology – CR CNIT Calculation

• Steps performed to calculate group combined reporting(CR) CNIT– Retrieve income and apportionment denominators from MN data– Create affiliated corporation EIN list– Match EIN list against PA data to determine filers– Retrieve separate CNIT and apportionment numerators from PA

data– Input above data into group spreadsheet to calculate tax (at 9.99%)– Calculate net impact as difference between CR CNIT and

aggregate separate company CNIT

6

Methodology – Population Calculation

• Greater than $1 billion– Because this group was completely sampled, there is no need to

blow up to the population• Less than $1 billion & Less than $1 million

– Total sample size – 123 groups– Total population size – 12,122 groups

• Net Impact Ratio– In order to determine the impact to the population, a net impact

ratio was calculated– [Net Impact from CR] / [MN Group Income]– Eliminates need to convert individual components (income,

apportionment, etc.)

Methodology – Regional Calculation

• The Minnesota data does not cover all corporationsthat are subject to PA CNIT

• Example – Electric Utilities• Regional Corporations

– To identify significant regional corporations, a list was made ofall corporations that have either

• $100 million in Pennsylvania sales• $1 million CNIT liability

– 651 corporations met this criteria, of which 437 werecovered in the MN data

7

Methodology – Regional Calculation

• Of the remaining 214 groups, 7 were electric utilities.• The impact for the other 207 was calculated using

sample data.– First, Pennsylvania income was converted to group income– “Flippers”– Group income was then multiplied by an aggregate net impact

ratio to arrive at net impact

Regional Calculation – “Flippers”

• Income of different signs reported to the states– Positive income in MN and negative income in PA (or vice versa)

• Using sample data, an income ratio (similar to netimpact ratio) was created along with a probabilityfigure.

• For example, a corporation that reported a loss to PAwould have a 47% chance of having reported positiveincome to MN.

• There is no way to determine which corporations arelikely to “flip”.– Monte Carlo simulation

8

Industry 60% Sales 75% Sales 100% Sales

Agriculture, Forestry, and Fishing (11.5) (11.2) (10.6)

Mining and Construction 2.3 2.2 1.9

Manufacturing 146.4 112.5 55.9

Transp., Commun., and Utilities 48.3 46.3 43.0

Trade 115.5 112.6 107.7

Finance, Insurance, and Real Estate 23.9 25.3 27.6

Services 16.3 14.9 12.6

Miscellaneous 24.2 26.7 30.9

Other 45.6 45.6 45.6

Total 411.0 374.8 314.5

Combined Reporting Estimate - Tax Year 2000

$2 million cap on separate NOLs & Uncapped Group NOLs @ 9.99%

Industry Disallowed $20 million cap Uncapped

Agriculture, Forestry, and Fishing (11.5) (11.5) (11.5)

Mining and Construction 3.0 2.6 2.3

Manufacturing 157.3 149.4 146.4

Transp., Commun., and Utilities 52.9 50.7 48.3

Trade 130.6 122.6 115.5

Finance, Insurance, and Real Estate 29.1 24.5 23.9

Services 28.6 22.8 16.3

Miscellaneous 28.6 25.3 24.2

Other 56.5 50.1 45.6

Total 475.1 436.5 411.0

Combined Reporting Estimate - Tax Year 2000

60% sales factor and $2 million cap on NOL Carry-ins

9

SIC Group CNIT

CR

Impact* Rate Cut New CNIT

Overall

Impact

Agriculture, Forestry, and Fishing 4.6 0.8 (1.1) 4.3 (0.2)

Mining and Construction 51.4 4.0 (11.1) 44.3 (7.0)

Manufacturing 341.8 119.2 (92.3) 368.7 26.9

Transportation, Communication, and Utilities 511.3 93.7 (121.1) 483.9 (27.4)

Trade 332.8 146.5 (96.0) 383.4 50.6

Finance, Insurance, and Real Estate 133.3 (12.7) (24.1) 96.4 (36.9)

Services 240.5 60.2 (60.2) 240.5 0.0

Miscellaneous 84.4 (1.8) (16.5) 66.1 (18.3)

1,700.0 410.0 (422.4) 1,687.6 (12.4)

* - CR Impact includes mandatory combined reporting at 7.99% with a single sales factor,

uncapped group NOLs, and market-based sourcing of sales for apportionment purposes.

Calculation of Overall Revenue Impact (Governor's Proposal)

Tax Year 2005

Conclusions

• House Bill 1557– Governor’s proposal– Additional Changes

• Tax Haven language– Referred to Finance Committee – May 2005

• Future Projects– Additional years of data from MN– Use data from another CR state to compare estimates

The Revenue Impact of Combined Reporting on the Corporate Net Income Tax

Pennsylvania Department of Revenue

FTA Research Award Nomination February 25, 2005

On March 4, 2004, Governor Rendell established a Business Tax Reform Commission

(BTRC) in Pennsylvania. The goal of the Commission was to create a more competitive business climate leading to greater economic growth as well as to ensure greater fairness in business taxation. In order to improve Pennsylvania’s competitive position, the Governor directed the Commission to evaluate the Commonwealth’s current business tax structure and recommend changes that would broaden the tax base, thus allowing for a corporate tax rate reduction while protecting the stability of the state budget.

The Commission considered the issues related to Pennsylvania’s system of separate company reporting of the corporate net income tax (CNIT). Separate company reporting uses a narrow tax base and allows tax-planning opportunities such as the use of passive investment companies (PICs) to shift income outside the Commonwealth. The Commission’s final report recommended the implementation of a mandatory unitary combined reporting system, which would create a tax base less susceptible to manipulation by requiring members of a unitary group to combine their income and expenses for tax purposes.

According to the Governor’s Executive Order establishing the BTRC, the Commission’s recommendations were required to be revenue neutral. Therefore, the Pennsylvania Department of Revenue analyzed the potential revenue impacts associated with implementing a unitary combined reporting requirement for the CNIT in conjunction with the other proposals recommended by the Commission, including:

• A reduction in the CNIT rate from 9.99% to 6.99%; • An increase in the sales factor weight from 60% to 100% for CNIT

apportionment purposes; • Uncapping NOLs generated by the combined group going forward; • Capping NOLs accrued prior to combined reporting at $2 million (consistent

with current law) and calculating and applying these NOLs on a separate company basis;

• Adoption of market-based sourcing for services for sales factor apportionment purposes; and

• Adoption of a 1% net income on pass-through entities.

In the 2005-06 Executive Budget, Governor Rendell proposed a business tax reform package that adopted all of the Commission’s recommendations with the exception of the 1% net income tax on pass-through entities. In order to maintain revenue neutrality, the Governor proposed reducing the CNIT rate to 7.99% rather than 6.99% as proposed by the Commission.

The recommendation of the BTRC and the Governor to adopt a mandatory unitary combined reporting system in Pennsylvania was based on the Department’s in-depth analysis of the revenue effects of combined reporting on CNIT revenues, including the related policy issues such as the NOL and apportionment rules. Section I of this paper outlines the methodology used by the Department to estimate the effect of combined reporting on CNIT revenues. Section II presents the Department’s estimates of the BTRC recommendations related to combined reporting.

I. Combined Reporting Estimate Methodology

A. Combined Reporting Data Sources

Because Pennsylvania is a strict separate entity reporting state, sufficient internal data necessary to measure the effect of combined reporting are not available. However, the Department receives information regarding corporations that file a Pennsylvania tax return from the Internal Revenue Service (IRS). The IRS provides the Department with a business master file (BMF), which contains federal tax return data for corporations that file a state return in Pennsylvania. In addition, the Bureau of Research requested a file of unitary combined group data from the state of Minnesota, which is a combined reporting state. Minnesota provided an electronic file of data for tax years 1999, 2000 and 2001.

Pennsylvania Tax System Data The Department of Revenue’s data tables indicate that there were 138,000 C corporations doing business in Pennsylvania that were subject to CNIT in 2000. Tax year 2000 data is the most complete data set currently available for use. The Pennsylvania data were used to calculate the current separate company CNIT liabilities for all unitary group members filing in Pennsylvania. In addition, Pennsylvania apportionment data were used in the numerator of the apportionment fraction for the unitary group. IRS BMF Data

At the federal level, most corporations that are members of an affiliated group elect to file a consolidated tax return for federal tax purposes. The federal tax returns of corporations that are not members of an affiliated group (usually smaller companies) would contain information related only to that one company. The 138,000 Pennsylvania C corporations in the Department’s data set were matched to the federal BMF data by employer identification number (EIN) and federal form 1120 line 28 income, which is the starting point in computing Pennsylvania taxable income. Out of the 138,000 C corporations, there were 63,500 whose federally reported EINs and incomes matched exactly the EINs and income figures reported on the Pennsylvania tax return. Because Pennsylvania is a strict separate entity reporting state, this exact match of EINs and incomes indicates that there were approximately 63,500 corporations filing in Pennsylvania that were not part of a federal consolidated group. In other words, they are mostly separate entities having no

2

affiliates at both the federal and state levels, and their tax liabilities would not be affected by changing to a combined reporting requirement. Of the 138,000 C corporations, 74,500 did not match to the BMF by both EIN and income. In other words, these corporations could have matched the BMF data by EIN and not income (i.e. reported a different income amount to Pennsylvania than was reported federally) or they did not match the BMF by EIN or income. For those companies that match by EIN but not income, the difference between the income amounts indicates that the federal tax return was filed on a consolidated basis to include a parent corporation and its affiliates, while the Pennsylvania tax return was filed for only one company (separate entity) in that group. Those companies that do not match by EIN or income may be members of a consolidated group. Therefore, these 74,500 corporations represent the population of corporations that are most likely to be members of a unitary business engaged in by more than one member of a combined group.

Minnesota Corporate Tax Data Minnesota provided combined income and apportionment data for all companies that apportion their income in the state for tax years 1999, 2000, and 2001. The Pennsylvania DOR matched the 74,500 EINs to the Minnesota corporate tax data, which resulted in the identification of 6,472 combined unitary groups filing in Minnesota whose parent corporation or affiliate also filed in Pennsylvania on a separate company basis.

B. Sample Selection It was not feasible to simulate a combined report for each Pennsylvania CNIT payer. Consequently, sampling techniques were created in order to estimate the impact on the entire population. The two variables used to determine sample size and mix were Standard Industrial Classification (SIC) code and the absolute value of the difference between Pennsylvania separate company federal income and Minnesota combined group income. For tax year 2000, the Minnesota data contained 107 combined groups (matched to Pennsylvania by EIN) whose Minnesota combined group income was either greater than or less than (absolute value) Pennsylvania separate entity income by at least $1 billion. It was determined that, because of this large difference, the potential impact on the overall estimate, and the fact that a significant amount of CNIT is paid by relatively few corporations, these 107 combined groups would all be sampled. The Minnesota data contained 6,365 groups whose combined group income differed (absolute value) from Pennsylvania separate entity income (matched by EIN) by less than $1 billion. A stratified sample was created for these groups by taking into consideration the prevalence of the groups’ industry by standard industrial classification (SIC) in the Minnesota data and the absolute value difference between the Minnesota and Pennsylvania income. These two factors were weighted together, which resulted in a sample of 123 groups representing companies in all SIC classifications for the absolute value classes of less than $1 billion and less than $1 million, with more sample groups being chosen from SIC classes that had a greater overall income difference. For example, 3 sample groups were chosen from the SIC 0 class,

3

which represents agriculture, forestry and fishing, while 24 sample groups were chosen from SICs 2,3, which represent manufacturing. In addition, because Minnesota sent three separate tax years’ worth of data, the sample was matched across all three years. For example, the sample for all tax years included all corporations that had an absolute income difference greater than $1 billion in the 1999, 2000, and 2001 data. This process allowed the Department to determine the impact of combined reporting net operating losses across the three years.

C. Calculation of Combined Reporting CNIT for the Sample Companies

The Pennsylvania CNIT liability calculated for the simulated combined group was compared to the actual tax liabilities for all of the members of the group that filed as separate entities in Pennsylvania. The net tax difference (gain or loss) between the CNIT liability of a simulated combined group and the aggregate CNIT liabilities for all members of that group reported by the separate entities filing in Pennsylvania is the amount used in deriving the estimate. In order to simulate a Pennsylvania unitary combined report, the following steps were taken:

• Minnesota group income and apportionment data were retrieved for a specific combined group. Pennsylvania tax reports for a member of the group were reviewed to obtain federal tax return data showing all members of the federal consolidated group by EIN.

• The federal EINs were cross-matched against Pennsylvania corporate tax data to determine which members of the consolidated group were filing as separate entities in Pennsylvania.

• The actual apportionment factor numerators of the Pennsylvania filers were aggregated to arrive at the total Pennsylvania property, payroll, and sales attributable to the combined group. Based on a sample of companies used in the estimate, the property and sales apportionment factor numerators were adjusted downward by 3.27% to eliminate potential intercompany transactions.

• The actual property, payroll, and sales factor apportionment denominators of the combined group from the Minnesota data were used to calculate the group apportionment percentage.

• The Minnesota unitary combined income was used as the starting point to arrive at Pennsylvania taxable income for the simulated combined group.

• Actual Pennsylvania additions to and deductions from income for the group were considered to arrive at combined Pennsylvania income to be apportioned.

• The combined Pennsylvania apportionment percentage was multiplied by the combined Pennsylvania income to be apportioned to arrive at simulated Pennsylvania combined taxable income.

4

• Pennsylvania combined taxable income was taxed at the rate of 9.99%, and the simulated combined tax liability was netted against the actual aggregate separate company tax liabilities for all members of the simulated combined group.

• The difference (positive or negative) between the simulated combined tax liability and the actual aggregate separate entity liabilities represents the gain or loss attributable to combined reporting for the group.

• This methodology was used for all three tax years (1999-2001). Companies picked up in the sample in any tax year were added to the sample for the other tax years. In addition, The Department was able to estimate the effect of uncapping the combined groups’ NOLs in subsequent years.

D. Weighting the Sample Results to Estimate the Population Effects

As mentioned, the estimated CNIT gain or loss for the 107 groups with an income difference between Minnesota and Pennsylvania greater than $1 billion was directly calculated for each group. Ratios from the sample of 123 groups were used in order to calculate the combined reporting impact for the remainder of the Minnesota combined groups. As Table 1 shows, for each absolute value class, a separate ratio was calculated by dividing the net impact of combined reporting on the sample by the Minnesota income for the sample, resulting in the computation of somewhat of an effective tax rate on combined income. The ‘net impact’ is the tax liability under combined reporting less the aggregate separate company liabilities for all affiliates identified on the federal schedules that file in Pennsylvania. As such, it would reflect the overall income and apportionment effects of the combined report as compared to the separate company reports. This ratio was then multiplied by the Minnesota incomes of the population that were in the same absolute value class as the sample to arrive at the net impact of combined reporting for that class.

Table 1

Net Impact Weights Tax Year 2000 – Single Sales Factor, $2 Million NOL Cap on Carry-Ins, Uncapped Group NOLs

Group Absolute Value Class Income Class MN Income Net Impact Ratio Greater than $1 billion Positive 158,869,200,050 228,365,995 0.001437

Negative 5,196,531,368 (49,923,780) (0.009607)

Less than $1 billion Positive 63,436,525,330 87,002,491 0.001371 Negative (23,783,859,063) (55,106,762) 0.002317

Less than $1 million Positive 1,317,230,125 905,503 0.000687 Negative (945,596,868) (2,562) 0.000003

Note: The Greater than $1 Billion Absolute Value Class is not used as a weight because all members of that group were included in the sample.

5

E. Regional Pennsylvania Corporations with No Presence in Minnesota

The Minnesota data and estimate methodology does not capture the effect of combined reporting on all large C corporations operating in Pennsylvania. For example, substantial Pennsylvania CNIT payers such as electric utility companies, which have no presence in Minnesota, are not included in the estimates based on Minnesota data. Therefore, a method was created in order to identify and account for these companies in the estimate. First, the Pennsylvania corporation data tables were queried to identify C corporations that had either $100 million in Pennsylvania sales or a $1 million or greater Pennsylvania CNIT liability. A total of 651 C corporations met these criteria. Once these corporations were identified, their EINs and the EINs of their parents, as determined by referencing the federal affiliate schedules, were matched against the Minnesota data. This cross matching revealed that 437 of these C corporations (or their parents) were in the Minnesota data, so they were already accounted for in the estimate. The remaining 214 large C corporations were not included in the estimate based on the Minnesota data. Seven of the 214 groups were electric utilities, and they were each sampled to determine their combined reporting impact. The impact for each of these utilities was calculated using a slightly different methodology than described earlier. Because these companies are not in the Minnesota data, combined income and apportionment figures were determined by reviewing these groups’ consolidated financial statements. Overall, the average tax impact results from the Minnesota stratified random sample were applied to the remaining 207 regional companies. However, in order to maintain the variability of the data for these companies, a randomizing technique was used to assign each taxpayer to a category of positive or negative tax change. This simulation technique enabled the Department to use micro-level data to evaluate the effects of various NOL proposals in conjunction with combined reporting. Tables 2 and 3 show the ratios applied to each group and the probability assigned to each.

Table 2

Regional Corporation Ratios Tax Year 2000 – Single Sales Factor, $2 Million NOL Cap on Carry-Ins, Uncapped Group NOLs

Income Class Income

PA MN Count PA MN Ratio Probability Positive Positive 44 38,839,532,560 51,363,385,196 1.3225 0.8302 Positive Negative 9 343,454,590 (1,103,277,044)) -3.2123 0.1698

Negative Positive 31 (1,723,415,294) 13,353,992,270 -7.7486 0.4697 Negative Negative 35 (20,058,820,544) (22,988,416,064) 1.1461 0.5303

6

Table 3

Minnesota Income Class Net Impact RatioGreater than $0 0.001358 Less than or Equal to $0 0.002228

F. Various Tax Policy Scenarios

The Commission considered various policy options in determining how the combined reporting requirement should be implemented in Pennsylvania. To assist with the evaluation of policy options, the Department calculated combined reporting impacts using various assumptions. For example, the Commission could recommend keeping Pennsylvania’s current 60% sales factor weight or they could increase the weight to 75% or 100%. The Commission could recommend keeping Pennsylvania’s current $2 million cap on net operating losses or they could remove the cap on losses incurred by the group. Tables 4 and 5 show the changes in the estimate caused by changing the assumptions of the combined reporting estimate for tax year 2000. The Department’s estimates of various tax policy scenarios demonstrate the significant effect that these policy options could have on industries affected by combined reporting. For example, increasing the weight of the sales from 60% to 100% decreases the combined reporting revenue gains from the manufacturing sector from $146.4 million to $55.9 million.

Table 4

Combined Reporting Estimate - Tax Year 2000 Sales Factor Changes with $2 million cap on separate NOLs & Uncapped Group NOLs Industry 60% Sales 75% Sales 100% Sales Agriculture, Forestry, and Fishing (11.5) (11.2) (10.6) Mining and Construction 2.3 2.2 1.9 Manufacturing 146.4 112.5 55.9 Transp., Commun., and Utilities 48.3 46.3 43.0 Trade 115.5 112.6 107.7 Finance, Insurance, and Real Estate 23.9 25.3 27.6 Services 16.3 14.9 12.6 Miscellaneous 24.2 26.7 30.9 Other 45.6 45.6 45.6 Total 411.0 374.8 314.5

7

Table 5

Combined Reporting Estimate - Tax Year 2000 Group NOL Options with 60% sales factor and $2 million cap on NOL Carry-ins Industry Disallowed $20 million cap Uncapped Agriculture, Forestry, and Fishing (11.5) (11.5) (11.5) Mining and Construction 3.0 2.6 2.3 Manufacturing 157.3 149.4 146.4 Transp., Commun., and Utilities 52.9 50.7 48.3 Trade 130.6 122.6 115.5 Finance, Insurance, and Real Estate 29.1 24.5 23.9 Services 28.6 22.8 16.3 Miscellaneous 28.6 25.3 24.2 Other 56.5 50.1 45.6 Total 475.1 436.5 411.0

II. Combined Reporting Revenue Estimate Table 6 below presents the results of the Department’s analysis of the effects of combined reporting for the sample companies and the population as a whole. The table reflects the proposal in the Governor’s 2005-06 budget: combined reporting with 100% sales factor, a $2 million cap on NOL carry-ins, uncapped combined reporting NOLs, market-based sourcing of sales, and a CNIT rate of 7.99%.

Table 6

SIC Group

Separate Company

CNIT

Combined Reporting

Effect

Total CNIT at

9.99%Rate Cut

CNIT at 7.99%

Overall Impact

Agriculture, Forestry, and Fishing 4.6 0.8 5.4 (1.1) 4.3 (0.2)Mining and Construction 51.4 4.0 55.4 (11.1) 44.3 (7.0)Manufacturing 341.8 119.2 461.0 (92.3) 368.7 26.9 Transportation, Communication, and Utilities 511.3 93.7 605.0 (121.1) 483.9 (27.4)Trade 332.8 146.5 479.3 (96.0) 383.4 50.6 Finance, Insurance, and Real Estate 133.3 (12.7) 120.6 (24.1) 96.4 (36.9)Services 240.5 60.2 300.7 (60.2) 240.5 (0.0)Miscellaneous 84.4 (1.8) 82.6 (16.5) 66.1 (18.3) 1,700.0 410.0 2,110.0 (422.4) 1,687.6 (12.4)

These estimates for tax year 2005 reflect the overall effects of combined reporting on CNIT revenues once fully phased in. The Department’s estimate of the long-term effects of uncapping NOLs is based on Minnesota data for tax year 2001 that indicates that NOLs reduce corporate income tax liability by approximately 10.7%. Based on the Department’s estimates for tax years 1999 through 2001, it is estimated that the BTRC recommendations incorporated into

8

9

the Governor’s Executive Budget for 2005-06 would increase CNIT liabilities by approximately 24% at the current 9.99% tax rate. Based on the Department’s estimates, the CNIT rate could be reduced to 7.99% to make the Governor’s Budget proposal essentially revenue neutral.

Pennsylvania Business TaxReform CommissionRevenue Estimate Update

Brenda S. WarburtonBureau of Research

PA Department of Revenue

October 20, 2004

1

Overview

• Updated combined reporting estimates

• Business Benefits Tax (BBT) estimate

• Revenue neutral business tax reform scenarios

2

Combined ReportingUpdated Revenue Estimates

Used 3 years of Minnesota data – 1999, 2000, and 2001

Objectives: (1) Estimate the potential revenue gains from combined reporting over multiple years(2) Measure the effect of NOL carryforwards generated under combined reporting for the unitary group

3

Combined Reporting by Industry:1999 to 2001

Industry 1999 2000 2001Agriculture, Forestry, and Fishing 6.0 (11.5) 0.4Mining and Construction 3.1 3.0 3.8Manufacturing 191.0 157.3 177.9Transp., Commun., and Utilities 33.6 52.9 47.5Trade 126.8 130.6 150.5Finance, Insurance, and Real Estate 13.9 29.1 (20.8)Services 24.9 28.6 30.0Miscellaneous 34.6 28.6 12.8Other 63.5 56.5 45.6Total 497.4 475.1 447.6

Data reflect $2M NOL carry-in cap and 60% sales factor. No impact from group NOLs.

4

Combined Reporting: Comparison to Previous Estimate - Tax Year 2000

Industry May OctAgriculture, Forestry, and Fishing 8.7 (11.5)Mining and Construction 12.7 3.0Manufacturing 122.0 157.3Transp., Commun., and Utilities 59.5 52.9Trade 146.7 130.6Finance, Insurance, and Real Estate 86.2 29.1Services 80.8 28.6Miscellaneous 14.3 28.6Other 0.0 56.5Total 530.9 475.1

5

Combined Reporting by Industry: Baseline Scenario: 1999 to 2001

Industry 1999 2000 2001Agriculture, Forestry, and Fishing 6.0 (11.5) 0.4Mining and Construction 3.1 2.3 3.6Manufacturing 191.0 146.4 163.3Transp., Commun., and Utilities 33.6 48.3 46.9Trade 126.8 115.5 143.5Finance, Insurance, and Real Estate 13.9 23.9 (22.3)Services 24.9 16.3 26.8Miscellaneous 34.6 24.2 11.8Other 63.5 45.6 41.8Total 497.4 411.0 415.8

Estimate reflects $2M cap carry-in NOLs; 60% sales factor; uncapped group NOL carry-forwards

6

Sales Factor Effects by Industry Tax Year 2000

Industry 60% Sales 75% Sales 100% SalesAgriculture, Forestry, and Fishing (11.5) (11.2) (10.6)Mining and Construction 2.3 2.2 1.9Manufacturing 146.4 112.5 55.9Transp., Commun., and Utilities 48.3 46.3 43.0Trade 115.5 112.6 107.7Finance, Insurance, and Real Estate 23.9 25.3 27.6Services 16.3 14.9 12.6Miscellaneous 24.2 26.7 30.9Other 45.6 45.6 45.6Total 411.0 374.8 314.5

7

Group NOL Effects by IndustryTax Year 2000

Industry Disallowed $20 million cap UncappedAgriculture, Forestry, and Fishing (11.5) (11.5) (11.5)Mining and Construction 3.0 2.6 2.3Manufacturing 157.3 135.9 146.4Transp., Commun., and Utilities 52.9 50.7 48.3Trade 130.6 122.6 115.5Finance, Insurance, and Real Estate 29.1 24.5 23.9Services 28.6 22.8 16.3Miscellaneous 28.6 25.3 24.2Other 56.5 50.1 45.6Total 475.1 423.0 411.0

8

NOL Estimate Issues• How to account for the phase-in of the

uncapped group NOL carry-forwards?-First 3 years of data show NOL cost reducing revenue gains from proposal-Separate company NOL “bank” carry-in to 2005 is expected to exceed $100 billion-Oldest NOL carry-in available in 2005 will be from 1995-1998 losses will be the first that can be carried forward for 20 years-Use of carry-ins from separate company “bank” plus uncapped group NOLs causes a short-term “bubble” in NOL usage

9

Estimate Assumptions• Estimates assume limiting $2M NOL carry-ins to 10

years of carry-forward to avoid a higher tax rate in the first few years

• Combined group NOLs assumed to be carried forward for 20 years

• NOLs assumed to be shared within a group• Estimate based on MN NOL rules (no deductions

for dividends received and foreign source income)• Fully phased in uncapped group NOL carry-

forwards reduces tax liability by 10.7%• Estimate does not include bank and insurance

company affiliates

Business Benefits Tax (BBT)• Tax base includes compensation, earnings (distributed and

undistributed), interest, rents and royalties

• Imposed on all businesses, except financial institutions and insurers subject to specialty corporation taxes

• $100,000 base exemption eliminates tax for the smallest taxpayers

• CNIT credit for BBT paid by C corps

• PIT deduction for BBT paid by pass-through entities and sole proprietors

• Estimated revenues for tax year 2005 at a 1% tax rate total $1,778.0 million (net of CNIT and PIT losses)

An Evaluation of the Estimation, by the Pennsylvania Department of Revenue, of the Revenue Impact of Combined Reporting on the Corporate Net Income Tax

Submitted to:

The Pennsylvania Department of Revenue Prepared by:

Global Insight, Inc.

September 7, 2004

James Diffley Group Managing Director

US RegionalServices Global Insight, Inc. 800 Baldwin Tower Eddystone, PA 19022

(610) 490-2642 Copyright 2004 Global Insight Inc.

INTRODUCTION The Commonwealth of Pennsylvania imposes a Corporate Net Income Tax (CNIT) upon corporations doing business in the state. Under the CNIT, corporations account for income by filing on a separate entity basis. For tax purposes, the corporate entity files a return based only on its activity, notwithstanding its relationship as parent, subsidiary, or member of an affiliated group of corporations. Many tax experts and authorities in numerous states now believe that separate reporting is conducive to aggressive tax planning by corporate groups and results in reduced state corporate income tax revenues. The Pennsylvania Business Tax Reform Commission has considered a tax reform that would require the filing of combined tax returns, often termed unitary taxation. The Pennsylvania Department of Revenue (DOR) has estimated the revenue implications of such a reform. The purpose of this report is to evaluate the methodology used in preparing that revenue estimate and to advise as to its robustness. The effect of the reform is to broaden the reach of the CNIT by inclusion of the activities of related corporations currently not subject to Pennsylvania's taxation. This new income is, of course, apportioned to Pennsylvania through a calculation of its share of the larger entity and, as such, does not necessarily generate an increase in Pennsylvania's CNIT liability. Because current corporate taxpayers report income on a separate basis, the DOR does not have corporate tax accounting information available to directly compute the tax liability of its taxpayers under the reform proposal. In such circumstances, revenue estimators typically resort to a methodology based on an estimate of more general economic variables that can be expected to serve as proxies for the relevant tax variables. In the case of state corporate income taxes, however—due to the volatility of profits over time, their variation over industries, and especially their sensitivity to corporate structure and inter-company transactions—any such estimates are far too imprecise to be credible policy guides. The best strategy, and the one subscribed to by revenue estimators across the country, is to attempt to simulate directly the tax accounting changes to a representative panel of sample tax returns. The comparison of Federal tax returns with Pennsylvania returns would provide an imperfect basis for a revenue estimate; but the lack of relevant multi-state accounting necessary for the exercise, and the confluence of related and unrelated business income within a consolidated group of corporations, would add considerable error to any estimate so derived. DOR has devised a methodology based upon an alternative data set that is directly relevant to the problem at hand. That is, it sought to directly use the tax return information revealed by combined returns as filed in other states. Minnesota, in particular, already requires the type of combined reporting under consideration as business tax reform in Pennsylvania. It is the opinion of Global Insight that the availability of these returns, filed on a combined reporting basis, represent the best available source data to use in construction of a Pennsylvania revenue estimate. The remainder of this report will evaluate the use by DOR of this resource.

2

METHODOLOGY Minnesota is one of sixteen states that requires combined reporting for corporate income tax liability. The Minnesota economy is roughly one-half the size of Pennsylvania's. It is not dramatically different in industrial structure, although its manufacturing sector has a greater high-technology component, and its private educational and health services sectors are less significant. One variant of the methodology might have utilized the impact for Minnesota of combined reporting and extrapolated that to Pennsylvania. However, while offering some insight, the revenue implications would be very inexact owing to the importance of unique corporate relationships between each state and its neighbors and the rest of the U.S. economy. These relationships affect the apportionment of the unitary group's income and may vary systematically with geography, size, and industry structure. The DOR, however, wisely realized that the Minnesota sample could be utilized in ways that were more directly representative of Pennsylvania. DOR's investigation directly uncovered a sample of unitary groups for whom at least one member was a Commonwealth taxpayer and whom would be required to file a combined return under the proposed reform. This sample would then provide the basis for the simulated tax impact analysis. Combined reporting would only affect a fraction of Pennsylvania's corporations. Those corporations, whether doing business entirely in-state or in multiple states, are not affected unless they are related to foreign (out-of-state) corporations in a related line of business. (Foreign members of an affiliated group who are in unrelated lines of business would not be required to file as part of the unitary group). In developing a sample of taxpayers containing members who currently are Pennsylvania taxpayers and who would be required under the reforms to file as part of a larger combine group, DOR chose precisely those unitary groups in Minnesota that contained members subject to the Pennsylvania CNIT. It has thus developed a sample database of tax returns that are directly applicable to the desired Pennsylvania tax simulation. These groups would be actual filers under combined reporting in the state. Therefore, it is an excellent database. The logical next question to address is whether it is representative of the true "universe" of all combined groups that would be mandated to file. If so, the inferences and estimates drawn from it would be subject only to the usual statistical prediction errors that occur in any well-drawn random sample. To evaluate the degree to which the sample is representative, we consider the following factors. First, as the sample is derived from the set of firms that operate in both Minnesota and Pennsylvania, does this set constitute a representative sample of all multi-state unitary groups? Second, is the selected sample representative of this group? On the first point, Minnesota is the 16th largest state in the nation. In an era when multi-state corporations dominate the U.S. business landscape from the East to West Coast, it is reasonable to consider the match of Minnesota and Pennsylvania firms as representative of large U.S. firms active nationwide. This group—of 4,643 unitary groups—also constitutes a large share of corporate tax liabilities and payments in most states. This

3

group is the most important component of liability to simulate; it adds great confidence to the resulting revenue estimate that it is truly representative. In addition, there exist what the DOR refers to as regional groups: those multi-state (generally in the Northeast) groups of corporations who do not operate nationwide, but may be significant taxpayers who would not be represented at all in Minnesota and hence not subject to sampling. The DOR did estimate the contribution of these groups; we will return later to this part of the methodology. On the second factor, the sample was selected in an efficient stratified manner. The largest taxpayers of interest, those whose unitary income (as indicated by the Minnesota returns) exceeded Pennsylvania income by more than $1 billion, were all included. These 83 groups accounted for 57% of the total income difference. Those with income differences from $1 million to $1 billion were sampled at a lesser rate, with 56 of 3,311 chosen. Finally, the smaller firms were sampled at the rate of 13 of 1,249. Stratification was also based on industry class. Based on an overview of the distribution of corporate income across the sample, the choices of strata appear to be sufficient for a good estimate. This sampling methodology represents "best practice" in statistical sampling methodology. The sample itself was a large portion of the CNIT, constituting about one-quarter of the total state CNIT liability. An adjustment was made by DOR to account for the effect of inter-company transactions on Pennsylvania apportionment. The treatment of transactions between members of the unitary group needs to be considered in any move to a combined reporting requirement. Other states generally apply a wash rule to negate the impact of such transactions. DOR used a sample of publicly available corporate records to adjust the reported Pennsylvania apportionment factors downward. It is unclear how accurate (a 3.27% reduction in Pennsylvania sales and property numerators) such an estimate is, but we judge it to be reasonably conservative. The results of tax calculations under combined reporting were then tabulated directly for the sample. The sample weights were then used to construct an estimate of all Pennsylvania taxpayers. This also is standard, and best, practice in sampling methodology for this type of estimation. The DOR provided Global Insight, under strict confidentiality agreements and with corporate identifiers removed, the raw data of the match of Pennsylvania and Minnesota tax returns. The organization of the data was transparent, and the techniques used to cumulate the return information were sound and straightforward to reproduce. We have no doubt as to the veracity of the results obtained. DOR simulated the revenue impact of combined reporting under two scenarios, distinguished by the treatment of Net Operating Loss (NOL) deductions. The calculation from the database is straightforward. It is important to note that in each alternative, it was assumed that NOL carry forwards could only be used by the taxpayer who, on a separate company basis, earned them. If the final legislation implementing the proposal were to allow the sharing of NOLs among group members, the revenue impact could be significantly different.

4

DOR found, among Pennsylvania taxpayers, 208 corporations with either CNIT liability in excess of $1 million or Pennsylvania sales in excess of $100 million, who were not included in the Minnesota match. The latter criterion is especially important as over half of the corporations meeting this criterion had no Pennsylvania CNIT liability in 2000. To simulate how these taxpayers, termed regional groups, would be affected by combined reporting, it applied the average income and apportionment adjustments implied by the moderately sized corporations (with income difference less than $1 billion) in the matched sample. The estimate for this group, which represents almost 30% of the revenue impact, is the weakest major part of the methodology. In statistical terms, it is comparable to applying the sample results of the moderately sized national groups to regional corporations. There is no reason to think that this is a biased estimate of the impact, but it does serve to increase the level of uncertainty and statistical error. A significant part of this uncertainty was ameliorated by an additional treatment for utilities, which accounted just over 50% of the liability of this category. For that sector, publicly available information from SEC filings was used to calculate alternative apportionment factors, which replaced those from the sample when smaller. This had the conservative effect of reducing the estimated revenue under combined reporting ACKNOWLEDGEMENTS Global Insight acknowledges the valuable professional assistance of the DOR staff in preparing this evaluation. In addition to providing the sample database in a well-organized and documented medium, the staff answered numerous questions regarding the development of the methodology. The information and explanations offered greatly clarified many issues. Among the topics discussed were the treatment of unrelated business income, the selection of sample statistics appropriate for the regional groups, the matching process among the three (federal, Pennsylvania, and Minnesota) taxpayers, the role of Pennsylvania nexus, and inter-company transactions. RECOMMENDATIONS The stimulation procedure was, though efficient, based on a statistical sample. As such, there is some degree of sampling error that results. Corporate income tax liability is notoriously volatile, both across corporations and across time. The confidence of any revenue estimate improves if the sample size is increased. Gains to increasing the sampling rate from the year 2000 Minnesota returns would not be expected to be large. However, owing to the variability over time of corporate income, it is advisable that the DOR attempt to reproduce the results with the returns of another tax year. This would add greatly to statistical confidence in any case. Moreover, as the year 2000 represented a U.S. business cycle peak, it is especially important to evaluate the revenue implications

5

6

of the complex interlocking web of corporate relationships at an alternative point in the business cycle. The use of NOLs by corporate taxpayers is, for instance, highly cyclical. CONCLUSION Global Insight, upon extensive evaluation, finds that the methodology the DOR used in estimating the revenue impact of combined reporting under the CNIT is sound and represents best practice. It is recommended that the DOR, in order to reduce the statistical uncertainty of the estimates, expands its sample to include information from additional tax years. The methodology itself can and should be retained for this analysis.